Visible polarization image fusion with non-subsampled Shearlets

摘要

For the disadvantage of traditional target detection methods with low detection rate and difficulties in distinguishing the small temperature difference and camouflage targets, this paper presents a kind of visible polarization image fusion method using non-subsampled Shearlets transform. Firstly, we obtain four polarization status images by multi-detector camera, where the direction of polarization angle is 0° 、 45° 、 90° 、 135° separately. The Stokes vectors are calculated by polarization status images. Then, the extracted target polarization feature images and light intensity image are decomposed into several sub frequency bands by NSST with fine multi-scale decomposition characteristics. Meanwhile, the fusion coefficients are determined based on high-frequency energy window and low-frequency mean in the frequency domain. At last, the final fusion image is obtained after NSST inverse transform and target enhancement. Experimental results illustrate that the proposed approach could obtain better fusion images with rich details, high contrast, highlighting the polarization characteristic of the target to improve the ability of scene perception and target detection.